矿井车载视频图像稳像算法研究

发布时间：2018-06-30 09:59

本文选题：矿井视频监控 + 车载视频　；参考：《工矿自动化》2017年11期

【摘要】：针对矿井车载摄像系统拍摄的视频因含有前景运动目标及高噪声造成的全局运动矢量估计误匹配率高、实时性较差等问题,提出了一种基于ORB特征匹配与改进粒子滤波的矿井车载视频图像稳像算法。在运动矢量估计阶段,采用ORB算法提取图像特征点;采用基于图像块的连续3帧间差分法,联合时空一致性准则快速剔除前景运动区域的特征点;结合前景标记区域,对特征点位置进行初次筛选,对保留下来的背景特征点进行配准;利用仿射变换模型实现帧间运动矢量的估计。在运动滤波阶段,采用基于估计窗的实时粒子滤波算法滤除抖动分量,获得补偿参数。实验结果表明,该算法有效避免了前景运动目标对稳像精度的影响,且具有较快的处理速度。
[Abstract]:Aiming at the problems of high mismatch rate of global motion vector estimation caused by the existence of foreground moving target and high noise in the video recorded by the vehicle camera system in mine, the real time performance of the video is poor, and so on. An image stabilization algorithm based on Orb feature matching and improved particle filter is proposed. In motion vector estimation stage, Orb algorithm is used to extract image feature points, image block based continuous three frame difference method is used to quickly eliminate feature points of foreground moving region by using spatio-temporal consistency criterion, and foreground tagging region is combined. The location of feature points is first selected, the remaining background feature points are registered, and the motion vectors between frames are estimated by affine transformation model. In the phase of motion filtering, the real time particle filter algorithm based on estimation window is used to filter the jitter component and obtain the compensation parameters. Experimental results show that the proposed algorithm can effectively avoid the influence of foreground moving targets on image stabilization accuracy and has a faster processing speed.
【作者单位】：中国矿业大学信息与控制工程学院;徐州市公安局科技处;
【基金】：江苏省“六大人才高峰”高层次人才培养项目(2015-ZBZZ-009)
【分类号】：TD76;TP391.41

【相似文献】